Lecture Notes in Computer Science 3472

13 Real-Time and Hybrid Systems Testing 359
systems) dense values. Therefore, the algorithms needed for generating test
cases have to be chosen carefully to gain a meaningful and manageable test
suite.
• Test Monitoring Again, the important factor is time. For monitoring the
execution of test cases, not only inputs and outputs to and from the SUT
are relevant, also the time at that these inputs and outputs occur must be
logged adequately.
• Test Evaluation Test evaluation also depends on time. Not only the correct
order of inputs and outputs is needed for deciding if a test has failed
or passed, the correct timing is also a crucial factor. An unavoidable fuzziness
for measuring time must be considered. The same holds for measuring
dense-valued data with respect to hybrid systems.
• Test Driver For real-time systems, the test driver must be capable of
giving inputs at the correct time, i. e. it must be fast enough for the SUT.
For hybrid systems, the same holds for dense data values, e. g. differential
equations have to be processed fast enough to model the valuation of data
values over time. If analog, i. e. dense-valued, data is expected as input from
the SUT, this must be generated. The most important function of the test
driver for real-time and hybrid systems is therefore bridging the gap between
the abstract model and the corresponding implementation, i. e. the SUT. On
the one hand, it has to concretize the input from the model to the SUT. On
the other hand, it has to abstract the output from the SUT for the test
monitor and test oracle.
All in all, we can identify two main issues for testing real-time and hybrid systems:
• Building manageable and meaningful models from the SUT.
• Finding manageable and meaningful test suites.
Test oracle, test monitor, and test driver are more a challenge in implementation.
They have in common that they depend on the model and the test case
generation to work efficiently. Therefore, the focus in the following chapters is
on modeling and test case generation.
13.3 Model-Based Test Case Generation
The aim of model-based testing is to derive test cases from an application model
that is an abstraction of the real behavior of the SUT. As described in Section
13.2.2, time can be abstracted either as discrete or dense time for real-time
systems. Values are considered discrete. For hybrid systems, the time domain is
dense just as the value domain. There are approaches for modeling each of these
systems as well as test case generation algorithms based on these models. These
are described in the following.